As pointed out by DaveReidUK in #723, the Extended Mode S data from FR24 is garbage. I've seen "data integrity issues" before, but this takes the cake.
The question arises: Is there any way to salvage the data? I'm sure the cockpit data recorder and the tower tapes will be very informative, but I'm not holding my breath waiting for the authorities to release that data in .csv format.
FR24 claims "Overall trends in the data are correct as reported" but I beg to differ, as documented below. They also claim "Flightradar24 interprets this data using advanced scripts and historical comparative data" but evidently their Ouija board is missing a few letters.
In particular, consider the two records timestamped 2020-05-22 09:36:39Z.840 and 2020-05-22 09:36:45Z.880. Both are from FR24 receiver 1971. During this 6-second interval, the Mach number was unchanged, the IAS went up by 2%, and the TAS went down by 30%. Also the heading changed by 171 degrees, which is about ten times faster than a standard-rate turn. Also the OAT changed by 8 °C in six seconds. It's not a fluke in these two readings; multiple records from multiple receivers exhibit the same preposterous misbehavior.
Sorry, I do not consider these trends to be "correct as reported". Apparently the bit strings are being wildly misinterpreted. Many elementary cross-checks are not being performed.
They also claim that they need data from four receivers in order to calculate position. I'm pretty sure that's wrong too; three receivers plus pressure altitude should be sufficient.
So, bottom line: Is there any way of obtaining usable data?